I am very open to the Subramanian critique of Indian GDP growth — national accounts data can get tricky and it's worth being critical. But I'd be curious what the night light data say. My read is these seem to show large improvements, but I don't know

The idea of comparing GDP growth directly to changes in capital inputs like steel production etc. I do think is very odd methodologically, since by construction this is going to fail to pick up genuine productivity growth; let alone services growth

The issue with commodity price deflators seems reasonable, but it seems to me it should cut both ways? ie periods of declining vs increasing commodity prices should result in reverse biases.

Finally, even if we accept the *growth rates* of GDP are off; what pins down the levels? The exercise implicitly assumes the base year is correct and we are testing shifts in changes, but do we think the levels are right? After all those are just chained past growth rates too